CarND P4: Advanced Lane finding¶
The goals / steps of this project are the following:
- Compute the camera calibration matrix and distortion coefficients given a set of chessboard images.
- Apply a distortion correction to raw images.
- Use color transforms, gradients, etc., to create a thresholded binary image.
- Apply a perspective transform to rectify binary image ("birds-eye view").
- Detect lane pixels and fit to find the lane boundary.
- Determine the curvature of the lane and vehicle position with respect to center.
- Warp the detected lane boundaries back onto the original image.
- Output visual display of the lane boundaries and numerical estimation of lane curvature and vehicle position.
In [1]:
#importing packages relevant to this project
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import numpy as np
import os
import cv2
%matplotlib inline
Camera Calibration¶
1. Loading the images¶
A simple method to load the images is defined. This method is used to load both the calibration as well as the test images
In [2]:
def load_images(names_path, names_list):
images = []
for name in names_list:
images.append(cv2.imread(names_path + name))
return images
2. Calibrating the camera¶
All the calibration images are loaded. The grid is a non conventional 9x6 grid. The Flowchart for Camera calibration is illustrated below.
In [3]:
# Path Containing Calibration images
calibration_path = 'camera_cal/'
# Grid Dimensions
nx = 9
ny = 6
# Generate a standard 9x6 grid
objp = np.zeros((ny*nx,3), np.float32)
objp[:,:2] = np.mgrid[0:nx,0:ny].T.reshape(-1,2)
calibration_image_names = os.listdir(calibration_path)
calibration_images = load_images(calibration_path,
calibration_image_names)
objpoints = []
imgpoints = []
for i in range(len(calibration_images)):
image = calibration_images[i]
image = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
ret, corners = cv2.findChessboardCorners(image, (nx,ny),None)
# If corners were found, add an instance of the objectpoints representing a grid
# Add Detected Corners to ImagePoints
if ret == True:
objpoints.append(objp)
imgpoints.append(corners)
# draw corners and show
img = cv2.drawChessboardCorners(image, (nx,ny), corners,ret)
In [4]:
# Compute the Calibration Matrices
ret, mtx, dist, rvecs, tvecs = cv2.calibrateCamera(objpoints, imgpoints, image.shape[::-1],None,None)
In [5]:
# Check for correct Calibration by testing on a reference checkerboard image
img = np.copy(calibration_images[0])
# undistort
dst = cv2.undistort(img, mtx, dist, None, mtx)
# Show undistorted and original images for verification
plt.figure(figsize=(20,10))
plt.subplot(221)
plt.xlabel('Original')
plt.imshow(img, cmap='gray')
plt.subplot(222)
plt.imshow(dst, cmap = 'gray')
plt.xlabel('Undistorted')
Out[5]:
In [6]:
# Test on the Test images
test_path = 'test_images/'
test_file_names = os.listdir(test_path)
test_images = load_images(test_path,test_file_names)
# Undistort all the test images and visualize the difference
for i in range(len(test_images)):
img = cv2.cvtColor(test_images[i],cv2.COLOR_BGR2RGB)
dst = cv2.undistort(img, mtx, dist, None, mtx)
plt.figure(i,figsize=(20,10))
plt.subplot(131)
plt.imshow(img, cmap='gray')
plt.xlabel('Original')
plt.subplot(132)
plt.imshow(dst, cmap = 'gray')
plt.xlabel('Undistorted')
plt.subplot(133)
plt.imshow(img-dst, cmap = 'gray')
plt.xlabel('Difference')
test_images[i]=dst
3. Calculating the Perspective Transform¶
The Inverse perspective transform is calculated using the four point method. An image is chosen deliberately with straight lanes so that the points are empirically tuned to get parallel lane lines and verify the IPM Visually.
In [7]:
# Create Source and Destination points for 4 point Perspective Transform
img = test_images[1]
img_size = img.shape
# upper and lower bounds for the height
height_ub = np.uint(img_size[0]/1.6)
height_lb = np.uint(img_size[0])
# Define points in perspective
top_scale_factor = 0.1
bottom_scale_factor = 0.95
center_point = np.uint(img_size[1]/2)
top_left = center_point - top_scale_factor*np.uint(img_size[1]/2)
top_right = center_point + top_scale_factor*np.uint(img_size[1]/2)
bottom_left = center_point - bottom_scale_factor*np.uint(img_size[1]/2)
bottom_right = center_point + bottom_scale_factor*np.uint(img_size[1]/2)
src = np.float32([[bottom_left,height_lb],
[bottom_right,height_lb],
[top_right, height_ub],
[top_left,height_ub]])
dst = np.float32([[30,img_size[0]+30],[img_size[1],img_size[0]+30],
[img_size[1],30],[30,30]])
M = cv2.getPerspectiveTransform(src, dst)
Minv = cv2.getPerspectiveTransform(dst,src)
plt.figure(figsize=(20,10))
plt.subplot(121)
plt.imshow(img,cmap='gray')
plt.scatter(x=src[:,0],y=src[:,1])
plt.subplot(122)
warped = cv2.warpPerspective(img, M,
(img.shape[1],img.shape[0]),
flags=cv2.INTER_LINEAR)
plt.imshow(warped,cmap='gray')
plt.scatter(x=dst[:,0],y=dst[:,1])
Out[7]:
In [8]:
# Load up test images to test the perspective transform calculated
ipm_images = []
for i in range(len(test_images)):
img = test_images[i]
warped = cv2.warpPerspective(img, M,
(img.shape[1],img.shape[0]),
flags=cv2.INTER_LINEAR)
plt.figure(figsize=(20,10))
plt.subplot(221)
plt.imshow(img,cmap = 'gray')
plt.xlabel('Perspective')
plt.subplot(222)
plt.imshow(warped,cmap='gray')
plt.xlabel('Top View')
ipm_images.append(warped)
4. Color space segmentation¶
HLS and LAB spaces were used to reliably separate lane markings based on color. The Flowchart is presented below
In [9]:
# Color Space Filter Pipeline
def color_pipeline(img):
# Convert to HLS
image_HLS = cv2.cvtColor(img,cv2.COLOR_RGB2HLS)
# Only use L Channel of HLS image
image_HLS = image_HLS[:,:,1]
# Define upper and lower bounds of L channel
hls_lb = 210
hls_ub = 255
# Create binary mask for HLS image
hls_mask = np.zeros_like(image_HLS)
hls_mask[(image_HLS > hls_lb) & (image_HLS <= hls_ub)] = 1
# Convert to LAB
image_LAB = image_HLS = cv2.cvtColor(img,cv2.COLOR_RGB2Lab)
# Only use B channel of LAB image
image_LAB = image_LAB[:,:,2]
# Define upper and lower bounds of B channel
lab_lb = 145
lab_ub = 255
# Create binary mask for LAB image
lab_mask = np.zeros_like(image_LAB)
lab_mask[(image_LAB > lab_lb) & (image_LAB <= lab_ub)] = 1
# Combine the masks
color_mask = np.zeros_like(image_LAB)
color_mask[(lab_mask == 1) | (hls_mask == 1)] = 1
return color_mask
5. Gradient based Segmentation¶
A sobel filter with x gradient only was used. The gradient based segmentation is to be performed on the top view (perspective transformed ) images so only a magnitude gradient in the x direction is relevant to the problem.
In [10]:
# Gradient Threshold Filter Pipeline
def gradient_pipeline(img):
# Convert to Grayscale for sobel filter
img_gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
# Since we are operating on the Warped images from top view,
# we can only work with x gradient magnitude
# Find X gradient
sobel_x = cv2.Sobel(img_gray, cv2.CV_64F, 1, 0, ksize=5)
# Get absolute value
abs_sobelx = np.absolute(sobel_x)
# Scale it from 0 to 255
scaled_sobelx = np.uint8(255*abs_sobelx/np.max(abs_sobelx))
# Define lower and upper bounds for the X gradient
xgrad_lb = 20
xgrad_ub = 255
# Create binary Mask
gradient_mask = np.zeros_like(scaled_sobelx)
gradient_mask[(scaled_sobelx > xgrad_lb) & (scaled_sobelx <= xgrad_ub)] = 1
return gradient_mask
6. Combination of Masks¶
The Masks are combined using a simple or combination. The underlying assumption is that either gradient or Color space segmentation should be able to pick up the lane markers.
In [11]:
# Test both color and gradient based pipelines for segmentation
lane_masks = []
for image in ipm_images:
image = cv2.GaussianBlur(image, (5,5), 0)
color_mask = color_pipeline(image)
grad_mask = gradient_pipeline(image)
plt.figure(figsize=(30,15))
plt.subplot(141)
plt.xlabel('color space filter',fontsize=18)
plt.imshow(color_mask,cmap='gray')
plt.subplot(142)
plt.xlabel('XGradient',fontsize=18)
plt.imshow(grad_mask,cmap='gray')
plt.subplot(143)
lane_mask = np.zeros_like(color_mask)
lane_mask [ ( color_mask == 1 ) | (grad_mask == 1 )] = 1
plt.xlabel('color + gradient combo',fontsize=18)
plt.imshow(lane_mask,cmap = 'gray')
plt.subplot(144)
plt.xlabel('Affine Warp',fontsize=18)
lane_mask_in_perspective = cv2.warpPerspective(lane_mask, Minv, (image.shape[1], image.shape[0]))
plt.imshow(lane_mask_in_perspective,cmap = 'gray')
lane_masks.append(lane_mask)
plt.show()
In [12]:
# Visualize the histogram
img = lane_masks[np.random.randint(0,len(lane_masks))]
histogram = np.sum(img[np.int32(img.shape[0]/2):,:], axis=0)
plt.figure(figsize=(20,10))
plt.subplot(211)
plt.imshow(img,cmap='gray')
plt.subplot(212)
plt.plot(histogram)
Out[12]:
In [13]:
# Line Class to track lines
from collections import deque
class Line:
def __init__(self):
# was the line detected in the last iteration?
self.detected = False
# x values of the last n fits of the line
self.recent_xfitted = deque([], 8)
# average x values of the fitted line over the last n iterations
self.bestx = None
# polynomial coefficients for the most recent fit
self.current_fit = [np.array([False])]
# x values for detected line pixels
self.allx = None
# y values for detected line pixels
self.ally = None
def track_line(
self,
x,
y,
fit_polynomial,
xfit,
curv,
):
self.detected = True
self.allx = x
self.ally = y
self.current_fit = fit_polynomial
self.recent_xfitted.append(xfit)
self.bestx = np.mean(self.recent_xfitted, axis=0)
In [14]:
# Import everything needed to edit/save/watch video clips
from moviepy.editor import VideoFileClip
from IPython.display import HTML
7. Detect and track lanes¶
The lane mask created by the combination of the gradient and the color masks is used to extract lane points by using the sliding window based approach on the histogram. The frame is vertically divided into 12 parts. and a moving window approach as instructed in the project help materials.
In [15]:
# Find lane points
def lane_points_finder(binary_warped, minpix = 50, margin =100):
histogram = np.sum(binary_warped[np.int32(binary_warped.shape[0]/2):,:], axis=0)
# Find the peak of the left and right halves of the histogram
# These will be the starting point for the left and right lines
midpoint = np.int(histogram.shape[0]/2)
leftx_base = np.argmax(histogram[:midpoint])
rightx_base = np.argmax(histogram[midpoint:]) + midpoint
# Choose the number of sliding windows
nwindows = 12
# Set height of windows
window_height = np.int(binary_warped.shape[0]/nwindows)
# Identify the x and y positions of all nonzero pixels in the image
nonzero = binary_warped.nonzero()
nonzeroy = np.array(nonzero[0])
nonzerox = np.array(nonzero[1])
# Current positions to be updated for each window
leftx_current = leftx_base
rightx_current = rightx_base
# Create empty lists to receive left and right lane pixel indices
left_lane_inds = []
right_lane_inds = []
# Step through the windows one by one
for window in range(nwindows):
# Identify window boundaries in x and y (and right and left)
win_y_low = binary_warped.shape[0] - (window+1)*window_height
win_y_high = binary_warped.shape[0] - window*window_height
win_xleft_low = leftx_current - margin
win_xleft_high = leftx_current + margin
win_xright_low = rightx_current - margin
win_xright_high = rightx_current + margin
# Identify the nonzero pixels in x and y within the window
good_left_inds = ((nonzeroy >= win_y_low) & (nonzeroy < win_y_high) & (nonzerox >= win_xleft_low) & (nonzerox < win_xleft_high)).nonzero()[0]
good_right_inds = ((nonzeroy >= win_y_low) & (nonzeroy < win_y_high) & (nonzerox >= win_xright_low) & (nonzerox < win_xright_high)).nonzero()[0]
# Append these indices to the lists
left_lane_inds.append(good_left_inds)
right_lane_inds.append(good_right_inds)
# If you found > minpix pixels, recenter next window on their mean position
if len(good_left_inds) > minpix:
leftx_current = np.int(np.mean(nonzerox[good_left_inds]))
if len(good_right_inds) > minpix:
rightx_current = np.int(np.mean(nonzerox[good_right_inds]))
# Concatenate the arrays of indices
left_lane_inds = np.concatenate(left_lane_inds)
right_lane_inds = np.concatenate(right_lane_inds)
# Extract left and right line pixel positions
leftx = nonzerox[left_lane_inds]
lefty = nonzeroy[left_lane_inds]
rightx = nonzerox[right_lane_inds]
righty = nonzeroy[right_lane_inds]
return leftx, lefty, rightx, righty
In [16]:
# Track lane points: To be used if last frame lane detection was successful
def lane_points_tracker(binary_warped, left_fit, right_fit, minpix = 50, margin =100):
nonzero = binary_warped.nonzero()
nonzeroy = np.array(nonzero[0])
nonzerox = np.array(nonzero[1])
margin = 100
# Use known fits to generate new points
left_lane_inds = ((nonzerox >
(left_fit[0]*(nonzeroy**2) +
left_fit[1]*nonzeroy +
left_fit[2] - margin)) &
(nonzerox <
(left_fit[0]*(nonzeroy**2) +
left_fit[1]*nonzeroy +
left_fit[2] + margin)))
right_lane_inds = ((nonzerox >
(right_fit[0]*(nonzeroy**2) +
right_fit[1]*nonzeroy +
right_fit[2] - margin)) &
(nonzerox <
(right_fit[0]*(nonzeroy**2) +
right_fit[1]*nonzeroy +
right_fit[2] + margin)))
# Again, extract left and right line pixel positions
leftx = nonzerox[left_lane_inds]
lefty = nonzeroy[left_lane_inds]
rightx = nonzerox[right_lane_inds]
righty = nonzeroy[right_lane_inds]
return leftx, lefty, rightx, righty
In [17]:
# Helper Functions for lane Extraction and fitting
# Function to get lane polynomials
def get_lane_polynomial(leftx, lefty, rightx, righty):
# Fit a second order polynomial to each set of points
left_fit = np.polyfit(lefty, leftx, 2)
right_fit = np.polyfit(righty, rightx, 2)
return left_fit, right_fit
# Function to get Lane polynomial for scaling to real world space
def get_scaled_lane_polynomial(leftx, lefty, rightx, righty):
ym_per_pix = 30/720 # meters per pixel in y dimension
xm_per_pix = 3.7/800 # meters per pixel in x dimension
# Fit new polynomials to x,y in world space
left_fit = np.polyfit(lefty*ym_per_pix, leftx*xm_per_pix, 2)
right_fit = np.polyfit(righty*ym_per_pix, rightx*xm_per_pix, 2)
return left_fit, right_fit
# Pad lane points by margin
def get_padded_lane_points(binary_image, left_fitx, right_fitx, margin = 100):
ploty = np.linspace(0, binary_image.shape[0]-1, binary_image.shape[0])
# Create points by padding with Margin
left_line_window1 = np.array([np.transpose(np.vstack([left_fitx-margin, ploty]))])
left_line_window2 = np.array([np.flipud(np.transpose(np.vstack([left_fitx+margin, ploty])))])
left_line_pts = np.hstack((left_line_window1, left_line_window2))
right_line_window1 = np.array([np.transpose(np.vstack([right_fitx-margin, ploty]))])
right_line_window2 = np.array([np.flipud(np.transpose(np.vstack([right_fitx+margin, ploty])))])
right_line_pts = np.hstack((right_line_window1, right_line_window2))
return left_line_pts, right_line_pts
# Get X Coordinates of lane points
def get_lane_xfits(binary_image, left_fit, right_fit, margin = 100):
# Generate x and y values for plotting
ploty = np.linspace(0, binary_image.shape[0]-1, binary_image.shape[0])
# Generate x points based on coefficients
left_fitx = left_fit[0]*ploty**2 + left_fit[1]*ploty + left_fit[2]
right_fitx = right_fit[0]*ploty**2 + right_fit[1]*ploty + right_fit[2]
return left_fitx,right_fitx
# Compute curvature based on polynomial
def get_curvature(binary_warped, left_fit, right_fit):
y_eval = np.float32(binary_warped.shape[0] - 1)
left_curverad = ((1 + (2*left_fit[0]*y_eval +
left_fit[1])**2)**1.5) / np.absolute(2*left_fit[0])
right_curverad = ((1 + (2*right_fit[0]*y_eval +
right_fit[1])**2)**1.5) / np.absolute(2*right_fit[0])
return left_curverad, right_curverad
# Get curvature in real world frame
def get_curvature_scaled(binary_warped, left_fit, right_fit):
y_eval = np.float32(binary_warped.shape[0] - 1)
# Define conversions in x and y from pixels space to meters
ym_per_pix = 30/720 # meters per pixel in y dimension
xm_per_pix = 3.7/800 # meters per pixel in x dimension
left_curverad = ((1 + (2*left_fit[0]*y_eval*ym_per_pix +
left_fit[1])**2)**1.5) / np.absolute(2*left_fit[0])
right_curverad = ((1 + (2*right_fit[0]*y_eval*ym_per_pix +
right_fit[1])**2)**1.5) / np.absolute(2*right_fit[0])
return left_curverad, right_curverad
# Get Distance from center
def get_distance_from_center(binary_image, left_fit, right_fit):
y_eval = np.float32(binary_image.shape[0] - 1)
# Define conversions in x and y from pixels space to meters
xm_per_pix = 3.7/800 # meters per pixel in x dimension
leftx = left_fit[0]*(y_eval)**2 + left_fit[1]*y_eval+ left_fit[2]
rightx = right_fit[0]*(y_eval)**2 + right_fit[1]*y_eval + right_fit[2]
center = binary_image.shape[1] / 2.
dl = (center - leftx) * xm_per_pix
dr = (rightx - center) * xm_per_pix
return dl, dr
# Check whether lane points are sane
def check_lane_sanity(left_curv, right_curv, left_fit, right_fit):
# check curvature
if left_curv >= 200 and right_curv >= 200:
flag_curv = True
else:
flag_curv = False
# check parallel
left_slope = left_fit[0]
right_slope = right_fit[0]
if np.absolute(right_slope - left_slope) <= 9e-4:
flag_paral = True
else:
flag_paral = False
return (flag_curv and flag_paral)
In [18]:
def lane_finder(binary_image, left, right):
# Set the width of the windows +/- margin
margin = 30
# Set minimum number of pixels found to recenter window
minpix = 50
# If lines are detected, only track
# If not Initialize or rely on best fit by history
if left.detected and right.detected:
(left_x, left_y, right_x, right_y) = \
lane_points_tracker(binary_image, left.current_fit,
right.current_fit, minpix, margin)
else:
(left_x, left_y, right_x, right_y) = \
lane_points_finder(binary_image, minpix, margin)
# Get Polynomial
(left_polynomial, right_polynomial) = get_lane_polynomial(left_x,
left_y, right_x, right_y)
# Get Scaled Polynomial
(left_fit_scaled, right_fit_scaled) = \
get_scaled_lane_polynomial(left_x, left_y, right_x, right_y)
# Get X co-ordinate
(left_xfit, right_xfit) = get_lane_xfits(img, left_polynomial,
right_polynomial)
# Get Curvature
(left_radius, right_radius) = get_curvature_scaled(img,
left_fit_scaled, right_fit_scaled)
# If lane lines are parallel and the curvature is within bounds
if check_lane_sanity(left_radius, right_radius, left_polynomial,
right_polynomial):
left.track_line(left_x, left_y, left_polynomial, left_xfit,
left_radius)
right.track_line(right_x, right_y, right_polynomial,
right_xfit, right_radius)
else:
left.detected = False
right.detected = False
ploty = np.linspace(0, binary_image.shape[0] - 1,
binary_image.shape[0])
# Create an image to draw the lane on
binary_color = np.zeros_like(binary_image).astype(np.uint8)
color_img = np.dstack((binary_color, binary_color, binary_color))
# Create an output image to draw lines on and visualize the result
out_img = np.dstack((binary_image, binary_image, binary_image)) \
* 255
window_img = np.zeros_like(out_img)
out_img = np.zeros_like(window_img)
# Pad lines to draw on lanes
(padded_lines_left, padded_lines_right) = \
get_padded_lane_points(binary_image, left.bestx, right.bestx)
points_left = np.array([np.transpose(np.vstack([left.bestx,
ploty]))])
points_right = \
np.array([np.flipud(np.transpose(np.vstack([right.bestx,
ploty])))])
# Draw the lane region
points = np.hstack((points_left, points_right))
cv2.fillPoly(color_img, np.int_([points]), (0, 255, 0))
# Draw the Lane Lines
cv2.fillPoly(window_img, np.int_([padded_lines_left]), (0, 255,
255))
cv2.fillPoly(window_img, np.int_([padded_lines_right]), (0, 255,
255))
# Combine the outputs
result = cv2.addWeighted(out_img, 1, window_img, 0.7, 0)
result = cv2.addWeighted(result, 1, color_img, 0.5, 0)
# Compute radius of Curvature in real world
y_eval = np.max(ploty)
ym_per_pix = 30 / 720 # meters per pixel in y dimension
xm_per_pix = 3.7 / 700 # meters per pixel in x dimension
# Calculate the new radii of curvature
left_curverad = (1 + (2 * left_fit_scaled[0] * y_eval * ym_per_pix
+ left_fit_scaled[1]) ** 2) ** 1.5 / np.absolute(2
* left_fit_scaled[0])
right_curverad = (1 + (2 * right_fit_scaled[0] * y_eval
* ym_per_pix + right_fit_scaled[1]) ** 2) ** 1.5 \
/ np.absolute(2 * right_fit_scaled[0])
# Get Distance from left and right from center
dl, dr = get_distance_from_center(binary_image,
left_fit=left.current_fit,
right_fit=right.current_fit, )
return result, left_curverad, right_curverad,dl,dr
In [19]:
# Global lines to use
left_line = Line()
right_line = Line()
# Function to process incoming video frame
def annotate_frame(image):
image = cv2.cvtColor(image,cv2.COLOR_BGR2RGB)
image_warped = cv2.warpPerspective(image, M,
(img.shape[1],img.shape[0]),
flags=cv2.INTER_LINEAR)
global left_line
global right_line
# Get Gradient Mask
gradient_mask = gradient_pipeline(image_warped)
# Get Color Mask
color_mask = color_pipeline(image_warped)
# Combine both
lane_mask = np.zeros_like(color_mask)
lane_mask [ (color_mask == 1) | (gradient_mask == 1) ] = 1
# Find lane
disp_image,left_curverad, right_curverad, dl, dr = lane_finder(lane_mask, left_line, right_line)
# Apply perspective transform
disp_image = cv2.warpPerspective(disp_image, Minv,
(image.shape[1],image.shape[0]),
flags=cv2.INTER_LINEAR)
# Add the result to the original image
result = cv2.addWeighted(image,1.0, disp_image,0.5,0)
# Annotate curvature on image
radText = 'Left_Radius:' + "%.10s" % (left_curverad) + " Right_Radius:" +"%.10s"%(right_curverad)
distText = 'Left_Distance:' + "%.10s" % (dl) + " Right_Distance:" +"%.10s"%(dr)
font = cv2.FONT_HERSHEY_SIMPLEX
cv2.putText(result, radText,(50,50), font, 1,(255,255,255),2)
cv2.putText(result, distText,(50,150), font, 1,(255,255,255),2)
# Show image for visualization
cv2.imshow('win', result)
cv2.waitKey(1)
return cv2.cvtColor(result,cv2.COLOR_RGB2BGR)
8. Test and Visualize the Lane Marking pipeline¶
The Entire Pipeline as described above, is tested on all the test images for the purpose of testing and tuning
In [20]:
# Test the pipeline
for image in ipm_images:
gradient_mask = gradient_pipeline(image)
color_mask = color_pipeline(image)
lane_mask = np.zeros_like(color_mask)
lane_mask [ ( color_mask >=0.5 ) | (gradient_mask >=0.5)] = 1
leftx, lefty, rightx, righty = lane_points_finder(lane_mask)
margin = 100
minpix = 50
# Get 2D polynomial
left_fit, right_fit = get_lane_polynomial(leftx, lefty, rightx, righty)
# Get X coordinates based on fit polynomial
left_fitx, right_fitx = get_lane_xfits(lane_mask,
left_fit,
right_fit,
margin)
# Get the Padded Lane Points using Margin
left_line_pts, right_line_pts = get_padded_lane_points(lane_mask,
left_fitx,
right_fitx,
margin)
out_image = np.dstack((lane_mask, lane_mask, lane_mask))*255
out_image[lefty, leftx] = [255, 0, 0]
out_image[righty, rightx] = [0, 0, 255]
disp_image = np.zeros_like(out_image)
cv2.fillPoly(disp_image, np.int_([left_line_pts]), (0,255, 255))
cv2.fillPoly(disp_image, np.int_([right_line_pts]), (0,255, 255))
result = cv2.addWeighted(out_image,1.0, disp_image,0.5,0)
result = cv2.cvtColor(result,cv2.COLOR_BGR2RGB)
plt.figure(figsize=(20,10))
plt.subplot(121)
plt.imshow(image,cmap='gray')
plt.xlabel('Original Image')
plt.subplot(122)
plt.imshow(result, cmap='gray')
plt.xlabel('Lane Markings')
9. Working with Video¶
The Pipeline after proper testing and tuning was used to annotate the video. The Flowchart is presented below
In [21]:
lane_output = 'project_annotated.mp4'
clip1 = VideoFileClip("./project_video.mp4")
annotated_clip = clip1.fl_image(annotate_frame)
%time annotated_clip.write_videofile(lane_output, audio=False)
In [22]:
cv2.destroyAllWindows()
cv2.waitKey(1)
Out[22]:
In [ ]: